Method for identifying factors influencing energy characteristics of vmd-pca subsynchronous oscillation

Abstract

The application utilizes variational mode decomposition (VMD) and principal component analysis (PCA), and proposes a VMD-PCA-based sub-synchronous oscillation energy characteristic influence factor identification method. The port energy characteristics of sub-synchronous oscillation under different working conditions are extracted through time domain simulation, and a regression model is established to perform regression analysis on the port energy of sub-synchronous oscillation. The process is as follows: the VMD mode decomposition is performed on the measured voltage and current at the outlet of the fan to obtain the voltage and current components under the sub-synchronous oscillation mode, and the transient energy flow at the outlet of the fan is obtained by calculation; the transient energy flow function of the fan outlet is fitted, and the energy flow power is taken as the stability characteristic quantity of sub-synchronous oscillation for research; the principal component method is used to solve the multicollinearity problem between the energy characteristic influence factors, a variable fitting evaluation model of the energy characteristic is established based on the stepwise regression method, and the key influence factors of the sub-synchronous oscillation energy characteristic are identified. The method can judge the current system energy characteristic of the system in time, identify the key influence factors affecting the system energy characteristic, and has important engineering significance for maintaining the safe and stable operation of the new power system.

Description

TECHNICAL FIELD

[0001] The application utilizes variational mode decomposition (VMD) and principal component analysis (PCA) to propose a VMD-PCA-based sub-synchronous oscillation energy characteristic influence factor identification method. In particular, it relates to a VMD-PCA-based sub-synchronous oscillation energy characteristic influence factor identification method caused by the grid connection of wind turbines in a wind farm through series compensation capacitors, and belongs to the field of power systems and their automation. BACKGROUND

[0002] With the development and utilization of large-scale wind farms and the penetration of high-proportion power electronic devices, the structure, operation characteristics and control mode of power systems have undergone fundamental changes, forming a new generation of new energy power systems. The mechanism and oscillation characteristics of sub-synchronous oscillation of power systems caused by the connection of wind farms are quite different from those of traditional thermal power units. The interaction between power electronic devices and between power electronic devices and the grid can easily trigger sub-synchronous oscillation, and the power system is at risk of instability.

[0003] Existing research, such as D-PMSG through LCC-HVDC transmission system sub-synchronous oscillation characteristic analysis by Gao Benfeng, Cui Yicen, Li Yunhong, etc. and impedance modeling and oscillation mechanism analysis of direct-drive wind farm through LCC-HVDC transmission system by Liu Bin, Husi, Wang Jiajun, etc., mostly uses state space modeling to solve characteristic roots or establishes impedance models to analyze equivalent impedance characteristics to analyze sub-synchronous oscillation characteristics. However, both state space method and impedance model require detailed modeling of the system under study, require obtaining the corresponding parameters of each element, and the system model requires a high order, which is time-consuming to calculate, and is mostly used for offline analysis.

[0004] Energy function construction method for AC-DC hybrid system by Liu Ye, Shen Chen, and Chen Houhe, Wang Changjiang, Jiang Tao, etc. based on port energy transient stability assessment of AC-DC hybrid system containing VSC-HVDC proposed by the author, etc. utilize power system measurement data to propose an energy function based on port energy. The construction process does not need to consider the dynamic process of the wind turbine, effectively avoiding the modeling process of the complex characteristics of power electronic links, and can fully utilize the measurable electrical quantities at the wind turbine port, reducing the difficulty of energy function construction.

[0005] Therefore, the application considers using VMD modal extraction technology to obtain port energy characteristics of subsynchronous oscillation under different working conditions, and establishes a regression model based on PCA dimension reduction method to perform regression analysis on subsynchronous oscillation energy, so that the current system energy characteristics of the system can be judged in time, the key influencing factors affecting the system energy characteristics are identified, and corresponding measures are taken when the system occurs subsynchronous oscillation, so as to maintain the safety and stability of the new power system. SUMMARY

[0006] The application proposes a subsynchronous oscillation energy characteristic influence factor identification method based on VMD-PCA in view of the deficiency of the current large-scale wind farm grid-connected subsynchronous oscillation transient energy flow propagation law research, wherein the VMD modal extraction technology and the regression analysis by establishing a regression model are key technologies for subsynchronous oscillation energy regression analysis, and the method is characterized by comprising the following steps:

[0007] (1) performing VMD modal decomposition on the voltage and current measured at the outlet of the wind turbine to obtain voltage and current components under subsynchronous oscillation mode;

[0008] (2) calculating the transient energy flow at the outlet of the wind turbine, performing linear fitting on the transient energy flow, obtaining the first derivative of the transient energy flow, that is, the energy flow power, and taking the energy flow power as a stable characteristic quantity of subsynchronous oscillation for research;

[0009] (3) performing principal component analysis on the independent variables affecting the energy flow power to reduce dimensions, obtaining a new sample set containing sufficient information of the original sample set, and performing stepwise regression analysis on the sample set after dimension reduction to establish a multiple linear regression model;

[0010] (4) obtaining the correlation between the original independent variables and the energy flow power based on the regression model, quantitatively analyzing the linear correlation between each influence factor and the energy flow power according to the regression coefficient, and then extracting the key influence factors of the energy flow power.

[0011] Beneficial effects:

[0012] The wind farm wind turbine control parameters and the AC line parameters are set, the three-phase voltage and the three-phase current under the subsynchronous oscillation frequency are obtained based on the VMD modal extraction technology, the subsynchronous oscillation transient energy flow is calculated, the energy flow power is obtained by linear fitting as a subsynchronous oscillation stability characteristic quantity, the principal component-stepwise regression method is used to establish a variable fitting evaluation model of energy characteristics, and the key influence factors affecting the subsynchronous oscillation energy characteristics are identified. The method is not limited to a specific system operation mode and unit control strategy, and the energy characteristic fitting and key factor analysis only depend on the measurable electrical quantities in the actual power system, and has a certain engineering application prospect. BRIEF DESCRIPTION OF DRAWINGS

[0013] Figure 1It is a VMD-PCA-based sub-synchronous oscillation energy characteristic influence factor identification method schematic diagram;

[0014] Figure 2 It is a wind farm grid-connected equivalent circuit diagram;

[0015] Figure 3 It is a wind turbine RSC control strategy schematic diagram of a wind farm;

[0016] Figure 4.1 It is a VMD decomposition IMF modal component time domain schematic diagram, including: IMF1 component time domain schematic diagram; IMF2 component time domain schematic diagram; IMF3 component time domain schematic diagram; IMF4 component time domain schematic diagram; IMF5 component time domain schematic diagram; IMF6 component time domain schematic diagram;

[0017] Figure 4.2 It is a VMD decomposition IMF modal component frequency domain analysis schematic diagram; including: IMF1 component time domain schematic diagram; IMF2 component time domain schematic diagram; IMF3 component time domain schematic diagram; IMF4 component time domain schematic diagram; IMF5 component time domain schematic diagram; IMF6 component time domain schematic diagram;

[0018] Figure 5 It is a comparison diagram of the initial measurement signal and the modal component at the sub-synchronous oscillation frequency; wherein, Fig. (a) is the time domain waveform of the initial measurement signal; Fig. (b) is the time domain waveform of the sub-synchronous oscillation frequency modal component after VMD modal extraction;

[0019] Figure 6 It is a comparison diagram of the initial signal and the modal component at the sub-synchronous oscillation frequency; wherein, Fig. (a) is the frequency spectrum analysis result of the initial measurement signal; Fig. (b) is the frequency spectrum analysis result of the sub-synchronous oscillation frequency modal component after VMD modal extraction;

[0020] Figure 7 It is the quantitative relationship between each variable and the energy flow power. DETAILED DESCRIPTION

[0021] The present application provides a VMD-PCA-based sub-synchronous oscillation energy characteristic influence factor identification method, in order to make the purpose, technical scheme and effect of the present application clearer, the specific implementation scheme of the present application is described in detail below in combination with the drawings and examples. The specific examples described in the present application are only used to explain the present application, not to limit the present application.

[0022] A VMD-PCA-based sub-synchronous oscillation energy characteristic influence factor identification method, characterized in that the method comprises the following steps:

[0023] Step 1: Adjust the line series compensation capacitor, wind speed, line flow, fan control parameters, etc. within a certain range, and measure the three-phase voltage and three-phase current signals at the fan outlet. The voltage and current measured at the fan outlet are decomposed by VMD mode to obtain the voltage and current components under the subsynchronous oscillation mode;

[0024] The VMD mode decomposition method for the voltage and current measured at the fan outlet is as follows:

[0025] (1) The VMD decomposition process can be regarded as a constrained variational problem:

[0026]

[0027] In the formula, f is the signal to be decomposed, u k is the kth IMF modal component obtained by VMD decomposition; w k is the center frequency of the kth IMF modal component; * is the convolution operation; is the derivative of the function with respect to time; δ(t) is the unit impulse function; t is the time; k represents the kth modal component obtained by VMD decomposition; j represents the imaginary unit.

[0028] (2) By introducing the Lagrange operator λ and the quadratic penalty factor σ, the constrained variational problem of step (1) is converted into an unconstrained variational problem, and the Lagrange equation of step (1) is obtained:

[0029]

[0030] (3) The alternating direction multiplier method is used to obtain the optimal solution of problem (2) through continuous iteration, and the iteration steps are as follows:

[0031] 1) Use random numbers to initialize the modal components and the center frequency

[0032] 2) Iteration and to obtain the optimal solution, is the n th iteration result of the center frequency of the k th IMF component, is the n th iteration result of the k th IMF component, is the n th iteration result of the Lagrange operator, and the formula is as follows:

[0033]

[0034]

[0035]

[0036] where τ is the noise tolerance of the signal, and performing inverse Fourier transform on u k (t), and the real part is the kth IMF component. and correspond to u i (t), f(t) and λ(t), respectively, k and i are the kth and ith IMF components, n+1 is the n+1th iteration, and w is the angular frequency introduced in the Fourier transform;

[0037] 3) Stop by repeated iteration until the convergence condition is met or the maximum number of iterations is reached, and the convergence condition is:

[0038]

[0039] where ε is the convergence criterion tolerance value, and the iteration is stopped when the above formula is met, and finally K IMF modal components {u1, u1,…, u K} are obtained.

[0040] Step 2: Calculate the transient energy flow at the outlet of the fan, linearly fit the transient energy flow, obtain the first derivative of the transient energy flow, i.e. the energy flow power, and use it as a stable characteristic quantity of the subsynchronous oscillation for research:

[0041] The transient energy flow at the outlet of the fan is calculated, the transient energy flow is functionally fitted, the first derivative of the transient energy flow, i.e. the energy flow power, is obtained, and it is used as a stable characteristic quantity of the subsynchronous oscillation for research, and the method is as follows:

[0042] (1) Considering that the port voltage and current electrical quantities are sinusoidal alternating values, under stable conditions, the sinusoidal function angular frequency is 2πf0, where f0=50Hz; under subsynchronous oscillation conditions, the wind farm has a subsynchronous oscillation source, which produces a sinusoidal wave with an angular frequency of 2πf er , where f er is the subsynchronous oscillation frequency, and based on the superposition theorem, the voltage and current electrical quantities at the port of the wind farm are composed of the power frequency sinusoidal wave and the subsynchronous frequency sinusoidal wave. Taking phase a as an example, the port voltage and current can be expressed as:

[0043] u a =U0cos(ω0t+δ0)+U er cos(ω er t+δ er )

[0044] i a =I0cos(ω0t+θ0)+I er cos(ω er t+θer )

[0045] where U0and I0are the RMS values of the fundamental frequency voltage and current, respectively; U er and I er are the RMS values of the subsynchronous frequency voltage and current, respectively; ω0= 2πf0, ω er are the angular frequencies of the fundamental frequency and subsynchronous frequency, respectively; δ0, θ0are the initial phases of the fundamental frequency voltage and current, respectively; δ er , θ er are the initial phases of the subsynchronous frequency voltage and current, respectively.

[0046] (2) After the Park transformation of the three-phase port voltage, the port voltage and current can be expressed as:

[0047]

[0048]

[0049] where ω - = ω0- ω er ; u d , u q are the port voltages in the dq coordinate system; i d , i q are the port currents in the dq coordinate system.

[0050] (3) The energy function of the branch L ij from node i to node j is as follows:

[0051]

[0052] where P ij and Q ij are the active power and reactive power of the branch L ij , respectively; i ij is the current of the branch L ij ; U i and θ i are the amplitude and phase angle of the voltage of node i, respectively; i d , i q are the branch currents of the branch L ij in the dq coordinate system; u d , u q are the node voltages of node i in the dq coordinate system; * is the conjugate symbol.

[0053] (4) Substituting the formula in (2) into (3), the node energy function can be expressed as:

[0054] W = ∫(i ddu q -i q du d )

[0055] =[-U er I0sin(ω_t-δ er +θ0)]-I er U er ω-cos(δ er -θ er )t

[0056] As can be seen from the above formula, after the system undergoes subsynchronous oscillation, the energy function consists of two parts: the component energy change term and the component energy consumption or generation term. The component energy change term is a periodic component, and the component energy consumption or generation term is a non-periodic component in the form of a linear function. Its linear coefficient can represent the damping characteristics of the component and is called energy flow power.

[0057] (5) Perform linear fitting on the transient energy flow:

[0058] W = ax + b

[0059] In the formula, the first derivative a of the transient energy flow function is the energy flow power, and b is the energy flow power constant term.

[0060] Step 3: Perform principal component analysis to reduce the dimensionality of the independent variables affecting energy flow power, obtain a new sample set containing sufficient information from the original sample set, perform stepwise regression analysis on the dimensionality-reduced sample set, and establish a multiple linear regression model.

[0061] The principal component analysis method is as follows:

[0062] (1) Standardized collection of raw indicator data: p-dimensional vector X = [x1, x2, ..., x...] p ], that is, p variables and n samples x i =[x i1 ,x i2 ,…,x ip ] T Let i = 1, 2, ..., n, n > p. Construct a sample matrix and perform a standardization transformation on the elements of the sample matrix, as shown in the following formula:

[0063]

[0064] in, Let j be the average value corresponding to the j-th variable. Let x be the variance of the j-th variable. ij Let X be the element in the i-th row and j-th column of the sample matrix X, from which the standardized matrix Z = [z1, z2, ..., zj] is obtained. p ];

[0065] (2) Calculate the correlation coefficient matrix R of the standardized matrix Z, and the formula is as follows:

[0066]

[0067] wherein, r ij is the element in the i-th row and the j-th column of the correlation coefficient matrix R, and z ij is the element in the i-th row and the j-th column of the standardized matrix Z;

[0068] (3) Solve the eigenvalues and eigenvectors of the matrix R, sort the eigenvalues according to the size, so that λ1≥λ2≥…≥λ p , and the corresponding orthogonal eigenvectors are V=[v1,v2,…,vp], to obtain the p principal components F of the sample X: p

[0069] F=VZ

[0070] F i =v 1i z1+v 2i z2+…+v pi z p ,i=1,2,…p;

[0071] (4) Calculate the variance contribution rate δ i , and determine the selected m principal components through the cumulative contribution rate α m , and the calculation formula of the variance contribution rate and the cumulative contribution rate is:

[0072]

[0073]

[0074] The stepwise regression analysis method is as follows:

[0075] (1) Establish a one-variable regression model Y=β0+β i X i +ε,i=1,2,…,m, for the m independent variables X1,X2,……,X m obtained by principal component analysis and the dependent variable Y, wherein ε is a random variable with a mean of 0, and the regression model is subjected to F test, the independent variable X i is the i-th factor affecting the energy function, and Y is the energy flow power that can represent the energy characteristics, and the calculation result is recorded as: Take the maximum value F I (1) , that is, For a given significance level α, record the corresponding critical value as F 1 , if F I (1) ​≥F 1 , then F I (1) Corresponding to the independent variable into the regression model, X I is the set of introduced variable indicators, X N is the set of non-introduced variable indicators, X Ik is the kth variable in the set of introduced variable indicators, X Nk is the kth variable in the set of non-introduced variable indicators.

[0076] (2) Establish the binary regression model of the dependent variable and the subset of independent variables {X I , X N1}, {X I , X N2}, …, {X I , X N(m-1)}, and calculate the value of the regression coefficient F test statistic, denoted as F k (2) , k = 1, 2, …, m-1, and select the maximum value as F I (2) , that is For a given significance level α, denote the corresponding critical value as F 2 , then F I (2) Corresponding to the independent variable into the regression model, update the set X I and X N , otherwise, terminate the variable introduction process.

[0077] (3) Repeat (2), each time selecting one from the independent variables X N that have not been introduced into the regression model, until no variable is introduced through testing. If L variables are finally selected, then the regression model can be obtained as:

[0078] Y = β0 + β1x1 + β2x2 + L + β L x L + ε

[0079] In the formula, β0, β1, β2, …, β L are regression coefficients, and ε is a random variable with a mean of 0, used to represent the influence of random factors on the dependent variable Y.

[0080] Step 4: Based on the regression model, obtain the correlation between the original independent variables and the energy flow power, quantitatively analyze the linear correlation between each influencing factor and the energy flow power according to the regression coefficients, and then extract the key influencing factors of the energy flow power:

[0081] The method for identifying key influencing factors of energy characteristics based on regression coefficients is as follows:

[0082] (1) Calculate F statistics and corresponding p-value to determine the feasibility of stepwise regression model;

[0083] (2) Obtain regression coefficients β0, β1, β2, …, β L from step 3, and quantitatively analyze the linear correlation between each influencing factor independent variable and energy flow power according to the regression coefficients, and then extract the key influencing factors affecting energy flow power.

[0084] Embodiments

[0085] Taking a wind farm grid-connected system as an example, the equivalent circuit diagram of the wind farm grid-connected system is shown in Figure 2 , and the RSC control system of the wind farm wind turbine is shown in Figure 3 . By adjusting the wind speed, series compensation capacitance, active power, reactive power, generator speed, voltage at the outlet of the wind turbine, and the inner loop proportional coefficient of the wind turbine RSC, the system generates subsynchronous oscillation.

[0086] 1. Adjust the wind farm parameters to obtain three-phase voltage and three-phase current data, and the adjustment range of some system parameters is shown in the following table. After VMD decomposition, multiple IMF components are obtained, as shown in FIG. 4, Figure 4.1 is the time-domain waveform of each IMF component, Figure 4.2 is the frequency domain analysis result of the IMF component, and by identifying the multiple modes decomposed, the subsynchronous oscillation mode component, i.e., IMF3, is obtained, Figure 5 (a) is the original time-domain signal without extraction, Figure 5 (b) is the time-domain waveform of the IMF3 mode component extracted. Figure 6 is the comparison diagram of the frequency distribution of the initial signal and the mode component at the subsynchronous oscillation frequency. It can be seen that the VMD extraction method can effectively extract the subsynchronous oscillation component.

[0087] Variable Range of variation Size of series compensation capacitor / μF 400～600 Wind speed / (m / s) 8～12 Equivalent wind farm rated capacity / MW 35～50 RSC current inner loop proportional coefficient 0.8～1.2

[0088] 2. Calculate the subsynchronous oscillation transient energy flow, and obtain the first derivative of the transient energy flow by function fitting. Part of the first derivative analysis results of the transient energy flow are shown in the following table:

[0089]

[0090] 3. Introduce principal component analysis method for dimensionality reduction analysis, project linearly related high-dimensional variables to low-dimensional principal component data, standardize the data in the data set, unify the data dimension, calculate the eigenvalues of the covariance matrix and the corresponding contribution rate quantification index, extract the principal components according to the contribution rate quantification index, and the contribution rate quantification index is shown in the following table:

[0091] Principal component Eigenvalue Variance contribution rate Cumulative variance contribution rate ​ 3.18137 45.44818 45.44818 Z2 2.14533 30.64763 76.09580 [Z3] 1.05766 15.10948 91.20528 [Z4] 0.39841 5.69162 96.89690 Z5 0.12322 1.76023 98.65712 Z6 0.06224 0.88914 99.54627 Z7 0.03176 0.45373 100.00000

[0092] From the table, the first three principal component eigenvalues are greater than 1, the fourth principal component eigenvalue is less than 1, and the cumulative variance contribution rate of the first three principal components is 91.20528%>90%. It is shown that the first three principal components basically contain all the information of the original signal and can reflect the main information contained in the seven variables. Therefore, the number of principal components is selected as three.

[0093] From the table, it can be seen that the principal component Z1 is mainly related to the wind speed x2, the active power fluctuation x3 at the outlet side of the wind farm, and the reactive power fluctuation x4 at the outlet side; the principal component Z2 is mainly related to the series compensation capacitance x1, the outlet side voltage fluctuation x5, and the generator speed x6; and the principal component Z3 is mainly related to the series compensation capacitance x1 and the RSC current inner loop proportional coefficient x7.

[0094] ​ Z2 [Z3] x1 -0.15871 -0.44122 -0.55630 x2 -0.50348 0.26686 -0.00620 x3 -0.49858 0.29239 -0.00991 x4 0.49649 -0.28743 -0.00999 x5 -0.31567 -0.49203 0.16469 x6 -0.33790 -0.48792 -0.14741 x7 0.11139 0.29299 -0.80090

[0095] 4. The principal components Z1, Z2, and Z3 are used as independent variables, and the energy function energy flow power is used as the dependent variable for regression analysis. Multiple linear regression and stepwise regression are performed respectively to obtain the regression model:

[0096] Y1=9.6269e-18+0.0440Z1+0.4920Z2-0.0129Z3

[0097] Y2=9.6269e-18+0.0440Z1+0.4920Z2

[0098] The regression model significance test quantity is shown in the following table:

[0099]

[0100]

[0101] By comparing the multiple linear regression and stepwise regression methods, it is found that after eliminating the principal component Z3, the F statistic value increases, and the p value corresponding to the F statistic decreases by one order of magnitude. The regression model is more accurate, so it is more effective to use stepwise regression to fit the variables of the energy characteristic. The regression coefficient can reflect the importance of each variable to the energy characteristic, thereby identifying the key influencing factors of the subsynchronous oscillation energy characteristic. The quantitative relationship between each variable and the energy flow power is shown in the following table. Figure 7

[0102] Figure 7 ​​It can be seen that the main influencing factors of energy flow power are series compensation capacitor x1, rotor side current inner loop proportional coefficient x7, generator speed x6 and voltage fluctuation x5. Among them, the rotor side inner loop proportional coefficient is positively correlated with the energy flow power, the larger the proportional coefficient, the greater the energy flow power, and the more unstable the system; series compensation capacitor x1, generator speed x6 and voltage fluctuation x5 are negatively correlated with the energy flow power, the greater the value, the smaller the energy flow power, and the more stable the system.

[0103] In summary, the method proposed in this paper can be independent of the complex establishment process of the system model, and is not bound to a specific system operation mode and unit control strategy. By analyzing the system subsynchronous oscillation energy characteristics fitting and quantitatively analyzing the influencing factors through system electrical quantity measurement data, the key influencing factors of the system energy characteristics are identified, and the method has certain engineering application prospect.

[0104] Finally, it should be noted that the above examples of the present application are merely examples for illustrating the present application, and are not intended to limit the embodiments of the present application. Although the applicant has described the present application in detail with reference to the preferred embodiments, those skilled in the art can make other different forms of changes and variations on the basis of the above description. It is impossible to exhaust all the embodiments here. Any obvious changes or variations derived from the technical solutions of the present application are still within the protection scope of the present application.

Claims

1. A method for identifying factors influencing energy characteristics of subsynchronous oscillation based on VMD-PCA, characterized in that, The method comprises the following steps: (1) VMD modal decomposition is performed on the voltage and current measured at the outlet of the fan to obtain voltage and current components under the subsynchronous oscillation mode; (2) The transient energy flow at the outlet of the fan is calculated, linear fitting is performed on the transient energy flow, the first derivative of the transient energy flow, i.e. the energy flow power, is obtained, and the energy flow power is taken as a stable characteristic quantity of the subsynchronous oscillation for research; (3) Principal component analysis is performed on the independent variables affecting the energy flow power, a new sample set covering sufficient information of the original sample set is obtained, stepwise regression analysis is performed on the reduced sample set, and a multiple linear regression model is established; (4) The correlation between the original independent variables and the energy flow power is obtained based on the regression model, the linear correlation between each influencing factor and the energy flow power is quantitatively analyzed according to the regression coefficients, and then the key influencing factors of the energy flow power are extracted; The VMD modal decomposition method of the voltage and current measured at the outlet of the fan in step (1) is as follows: (1) The VMD decomposition process can be regarded as a constrained variational problem: wherein f is the signal to be decomposed, u k is the kth IMF modal component obtained by VMD decomposition; w k is the center frequency of the kth IMF modal component; * is the convolution operation; is the time derivative of the function; δ(t) is the unit impulse function; t is time; k represents the kth modal component obtained by VMD decomposition; j represents the imaginary unit; (2) By introducing the Lagrange operator λ and the quadratic penalty factor σ, the constrained variational problem in step (1) is converted into an unconstrained variational problem, and the extended Lagrange equation in step (1) is obtained: (3) The optimal solution of the problem in step (2) is obtained through continuous iteration by using the alternating direction multiplier method, and the iteration steps are as follows: 1) Initialization of modal components with random numbers and center frequency 2) iteration and finding the optimal solution, is the nth iteration result of the center frequency of the kth IMF component, is the nth iteration result of the kth IMF component, is the nth iteration result of the Lagrange operator, which is given by where τ is the noise tolerance of the signal, and the real part of the inverse Fourier transform of k (t) is the kth IMF component; and correspond to the Fourier transform of i (t), f(t) and λ(t), k, i are the kth, ith IMF component, n+1 is the n+1th iteration, and w is the angular frequency introduced in the Fourier transform. 3) Stop through repeated iteration until the convergence condition is met or the maximum number of iterations is reached, and the convergence condition is: wherein ε is a convergence criterion tolerance value, and if the above equation is satisfied, the iteration is stopped, and finally K IMF modal components {u1, u1, …, u K} can be obtained. K} can be obtained. (4) identifying the K IMF modal components obtained by VMD decomposition, and finally obtaining the voltage u and current component I under the subsynchronous oscillation modal under subsynchronous oscillation abc and current component I abc ; The method for calculating the transient energy flow at the outlet of the fan, function fitting the transient energy flow, obtaining the first derivative of the transient energy flow, i.e. the energy flow power, and taking the energy flow power as a stable characteristic quantity of the subsynchronous oscillation for research in step (2) is as follows: (1) Considering that the port voltage and current are sinusoidal AC values, under steady-state conditions, their sinusoidal angular frequency is 2πf0, where f0 = 50Hz; under subsynchronous oscillation conditions, the wind field has a subsynchronous oscillation source, which will generate an angular frequency of 2πf0. er A sine wave, where f er Given the subsynchronous oscillation frequency, based on the superposition theorem, the voltage and current electrical quantities at the wind farm port are composed of the superposition of the power frequency sine wave and the subsynchronous frequency sine wave; taking phase a as an example, the port voltage and current can be expressed as: u a = U0cos(ω0t + δ0) + U er cos(ω er t + δ er ) i a = I0cos(ω0t + θ0) + I er cos(ω er t + θ er ) wherein U0 and I0 are the effective values of the power frequency voltage and power frequency current respectively; U er and I er are the effective values of the subsynchronous frequency voltage and subsynchronous frequency current respectively; ω0 = 2πf0, ω er are the angular frequencies of the power frequency and subsynchronous frequency respectively; δ0, θ0 are the initial phases of the power frequency voltage and power frequency current respectively; δ er , θ er are the initial phases of the subsynchronous frequency voltage and subsynchronous frequency current respectively; (2) After the three-phase port voltage is converted to the dq coordinate system through Park transformation, the port voltage and current can be represented as: where ω - = ω0- ω er ; u d , u q are the port voltages in dq coordinates; i d , i q are the port currents in dq coordinates; (3) from node i via branch L ij The energy function to flow to node j is given by where P ij and Q ij are the active and reactive power of branch L ij , i ij is the current of branch L ij , U i and θ i are the amplitude and phase angle of the voltage of node i; i d , i q are the branch currents in dq coordinate system on branch L ij ; u d , u q are the node voltages on node i in dq coordinate system; * is the conjugate symbol; (4) By substituting the formula in (2) into (3), the node energy function can be represented as: W = ∫(i d du q -i q du d ) = [-U er I0sin(ω_t-δ er +θ0)]-I er U er ω-cos(δ er -θ er )t From the above formula, it can be seen that after the system occurs subsynchronous oscillation, the energy function is composed of two parts, i.e. the element energy change term and the element consumption or generated energy term, wherein the element energy change term is a periodic component, and the element consumption or generated energy term is a non-periodic component in the form of a first-order function, and the coefficient of the first-order term can represent the damping characteristic of the element, which is called energy flow power; (5) Linear fitting is performed on the transient energy flow: W = ax + b In the formula, the first derivative a of the transient energy flow function is the energy flow power, and b is the constant term of the energy flow power; The principal component analysis method in step (3) is as follows: (1) Standardized collection of original index data p-dimensional vector X = [x1, x2, …, xp], that is, p variables, n samples x p = [x i 1, x i1 2, …, x i2 n] ip , i = 1, 2, …, n, n > p, construct a sample matrix, and perform a standardized transformation on the sample matrix elements, as follows: T (2) Calculate the correlation matrix R of the sample matrix X, and the formula is as follows: wherein, is the average value for the jth variable, is the variance of the jth variable, x ij is the element in the ith row and jth column of the sample matrix X, thus obtaining the standardized matrix Z = [z1, z2,..., z p ]. (2) The correlation coefficient matrix R of the standardized matrix Z is calculated, and the formula is as follows: wherein r ij is the element in the i-th row and j-th column of the correlation matrix R, z ij is the element in the i-th row and j-th column of the normalized matrix Z; (3) Solve the eigenvalues and eigenvectors of the matrix R, sort the eigenvalues by size, so that λ1≥ λ2≥…≥ λp p , the corresponding orthogonal eigenvectors are V = [v1, v2, …, vp] p , get the p principal components F of the sample X: F = VZ F i = v 1i z1+ v 2i z2+…+v pi z p i = 1, 2, … p; (4) Calculate the variance contribution rate δ i , by accumulating the contribution rate α m Determine to select m principal components, the variance contribution rate, the cumulative contribution rate calculation formula is: The stepwise regression analysis method in step (3) is as follows: (1) The m independent variables X1, X2, …, X m A regression model Y = β0+ β i X i + ε, i = 1, 2, …, m is established, where ε is a random variable with a mean of 0. The regression model is subjected to F test, the independent variable X i is the i th factor affecting the energy function, and Y is the energy flow power that can represent the energy characteristics. The calculation result is recorded as: Take the maximum value F I (1) That is For a given significance level α, record the corresponding critical value F 1 , if F I (1) ≥ F 1 , F I (1) Corresponding independent variable into the regression model, record X I as the index set of introduced variables, record X N as the index set of non-introduced variables, X Ik is the kth variable in the index set of introduced variables, X Nk is the kth variable in the index set of non-introduced variables; (2) Establish the binary regression model of dependent variable and the subset of independent variables {X I ,X N1}, {X I ,X N2}, …, {X I ,X N(m-1)}, calculate the value of the regression coefficient F test statistic of the variable, denoted as F k (2) , k = 1, 2, …, m-1, select the maximum value as F I (2) , that is For a given significance level α, denote the corresponding critical value as F 2 , then F I (2) Corresponding to the introduction of the regression model of the independent variable, update the set X I and X N , otherwise, terminate the variable introduction process; (3) Repeat (2) each time from the variables X not yet introduced into the regression model N one at a time until no variables are introduced by the test, and if finally L variables are selected, then the regression model is given by Y = β0+ β1x1+ β2x2+ L + β L x L + ε where β0, β1, β2,..., β L are regression coefficients and ε is a random variable with mean 0 representing the effect of random factors on the dependent variable Y.

2. The VMD-PCA-based subsynchronous oscillation energy characteristic influence factor identification method according to claim 1, characterized in that, The method for identifying the key influencing factors of the energy characteristic based on the regression coefficients in step (4) is as follows: (1) Calculate the F statistic and the corresponding p value to determine the feasibility of the stepwise regression model; (2) Regression coefficients β0, β1, β2, …, β L According to the regression coefficients, the influence factors of each variable on the power of the energy flow can be quantitatively analyzed.

3. A non-volatile storage medium, characterized by, The non-volatile storage medium comprises a stored program, wherein the program controls the device where the non-volatile storage medium is located to execute the method of claim 1 when the program is running.

4. An electronic device, characterized by comprising: A computer readable storage medium having stored computer readable instructions, wherein the computer readable instructions, when executed by a processor, perform the method of claim 1. A computer readable storage medium having stored computer readable instructions, wherein the computer readable instructions, when executed by a processor, perform the method of claim 1.

Images